I realise this is a 3-year-old thread that has been dredged up, but since I find floating-point numbers interesting (as a professional programmer), I thought I'd leave this link to a website I find useful: https://www.h-schmidt.net/FloatConverter/IEEE754.html
It demonstrates that in general floating-point numbers can almost never represent fractions exactly (see 'Value actually stored in float'). As a result, you will always be able to find edge cases where rounding is apparently being performed incorrectly.
© Espruino, powered by microcosm.
Report a problem